Detecting prolonged sitting bouts with the ActiGraph GT3X

Abstract

The ActiGraph has a high ability to measure physical activity; however, it lacks an accurate posture classification to measure sedentary behavior. The aim of the present study was to develop an ActiGraph (waist‐worn, 30 Hz) posture classification to detect prolonged sitting bouts, and to compare the classification to proprietary ActiGraph data. The activPAL, a highly valid posture classification device, served as reference criterion. Both sensors were worn by 38 office workers over a median duration of 9 days. An automated feature selection extracted the relevant signal information for a minute‐based posture classification. The machine learning algorithm with optimal feature number to predict the time in prolonged sitting bouts (≥5 and ≥10 minutes) was searched and compared to the activPAL using Bland‐Altman statistics. The comparison included optimized and frequently used cut‐points (100 and 150 counts per minute (cpm), with and without low‐frequency‐extension (LFE) filtering). The new algorithm predicted the time in prolonged sitting bouts most accurate (bias ≤ 7 minutes/d). Of all proprietary ActiGraph methods, only 150 cpm without LFE predicted the time in prolonged sitting bouts non‐significantly different from the activPAL (bias ≤ 18 minutes/d). However, the frequently used 100 cpm with LFE accurately predicted total sitting time (bias ≤ 7 minutes/d). To study the health effects of ActiGraph measured prolonged sitting, we recommend using the new algorithm. In case a cut‐point is used, we recommend 150 cpm without LFE to measure prolonged sitting and 100 cpm with LFE to measure total sitting time. However, both cpm cut‐points are not recommended for a detailed bout analysis.

Concurrent and discriminant validity of ActiGraph waist and wrist cut-points to measure sedentary behaviour, activity level, and posture in office work

“…Furthermore, if the measured times are interpreted on an individual level to detect daily behaviour patterns or prolonged sitting bouts [53], the simplification of equating minPA with sitting holds no longer true. In line with Matthews et al [4], we noticed in a recent study a very small bias (− 7 min per day) between minPA (ActiGraph, waist-worn, VA, 100 cpm) and sitting (activPAL), but a very large bias (− 105 min per day) between the two sensors when looking at the time spent in bouts ≥10 min [54]. A similar underestimation of prolonged sitting was found in other studies [52,55], indicating that prolonged sitting (activPAL) contains some LPA minutes (ActiGraph) breaking up SB.…”

How Accurate and Precise Can We Measure the Posture and the Energy Expenditure Component of Sedentary Behaviour with One Sensor?

“…To measure both components of sedentary behaviour with one sensor, the proprietary posture classification of the activPAL was combined with an energy expenditure algorithm to activPAL+, while the proprietary energy expenditure classification of the ActiGraph was combined with a posture algorithm to ActiGraph+. The development of the activPAL algorithm is presented here, while the development of the ActiGraph algorithm is presented in [ 15 ]. Both algorithm developments used the same data and development procedure.…”

“…In a previous study, the 75 cpm cut-point turned out to have a substantially higher validity (kappa of 0.69 compared to an indirect calorimeter) than the frequently used 100 cpm (kappa of 0.56) to separate inactive (≤1.5 METs) and active sitting (>1.5 METs) [ 21 ]. Subsequently, three different feature sets were generated: Feature Set 1 consisted of the same 563 signal features already used for the ActiGraph algorithm development [ 15 ]. Feature Set 2 consisted of the 213 time-based features for the raw sensor axis and vector magnitude of Feature Set 1.…”

“…The ActiGraph+ processing used the new ActiGraph posture algorithm [ 15 ] and classified each minute into sitting, standing, and locomotion and subsequently reclassified each sitting minute into sedentary behaviour and active sitting using the corresponding ActiGraph counts and a cut-point of 75 cpm.…”

“…Thus, the activPAL’s sedentary estimate is exclusively based on posture (i.e., sitting), while the ActiGraph’s sedentary estimate is exclusively based on energy expenditure (i.e., ≤1.5 METs). For both sensor types, efforts have been made to develop new algorithms to classify the other component, i.e., energy expenditure for posture sensors [ 11 , 12 ] and posture for movement sensors [ 13 , 14 , 15 ]. However, thus far, these algorithms have not been combined with the proprietary data processing of the sensors to classify posture and energy expenditure simultaneously.…”